Closer/seed/questions/QUESTION_QUALITY_CHECKLIST.md

6.0 KiB

Closer Question Quality Checklist v3

See also: QUESTION_CONTENT_GUIDE.md | QUESTION_SCHEMA.md | QUESTION_REWRITE_PLAN.md

Purpose

This checklist prevents technically valid but boring questions from reaching the app.

Passing JSON validation is not enough.

Every question must also feel human, useful, and worth answering.

Automatic Rejects

Reject any question that contains or strongly resembles:

  • Describe...
  • Reflect on...
  • Discuss...
  • Evaluate...
  • In what ways...
  • How satisfied are you...
  • What boundary around...
  • Explore your feelings...
  • Identify the ways...
  • Rate the effectiveness...
  • Communication style
  • Emotional processing
  • Conflict framework
  • Relationship dynamic

These are therapy worksheet patterns.

Burn them politely.

Daily Pack Hard Checks

For the daily single choice weekday pack, confirm before content review:

  • 500 total questions
  • 75 free questions
  • 425 premium questions
  • every question is single_choice
  • every question has exactly one weekday tag
  • every question has 4 to 6 options

Daily Pack Rejects

For daily single choice weekday questions, reject any prompt that feels like:

  • therapy homework
  • self-help content
  • an HR wellness survey
  • a communication worksheet
  • abstract emotional processing
  • generic AI relationship advice

Reject daily questions using these words or phrases unless there is a strong reason:

  • reset
  • process
  • mental load
  • emotional load
  • autopilot
  • pressure
  • soft landing
  • relationship dynamic
  • name the mood
  • emotional processing
  • communication style
  • conflict framework

Bad daily question:

Which reset would help us avoid autopilot?

Good daily question:

Long day. What would actually help tonight?

Bad daily question:

What would help us process the emotional load of the week?

Good daily question:

What tiny nice thing would make tonight better?

Daily Question Voice Test

A daily question passes only if the answer is yes to all of these:

  • Would a real couple answer this in under 10 seconds?
  • Does it sound like something normal people would say?
  • Is it about today or normal life?
  • Are the options concrete?
  • Does every option answer the prompt?
  • Is it warm, playful, honest, or useful?

Reject it if:

  • it sounds clinical
  • it sounds like advice
  • it sounds like therapy
  • it is too abstract
  • it uses a wellness app phrase
  • it could belong in any generic relationship app

Duplicate Pattern Rejects

Reject the question if it repeats too much of a previous question:

  • same opening phrase
  • same option list
  • same emotional goal
  • same situation
  • same answer pattern
  • same wording with one noun swapped

A file can be valid and still feel like a cursed spreadsheet.

Opening Variety

No category should overuse any opening.

Watch for overuse of:

  • What should we...
  • What do you...
  • Which...
  • How much...
  • What makes...

These are allowed, but not as the backbone of the pack.

Consumer Test

Ask:

Would a real couple answer this voluntarily on a Friday night?

Reject if the honest answer is no.

Conversation Test

A question should create at least one of these:

  • a laugh
  • a story
  • a decision
  • a date idea
  • a compliment
  • a surprise
  • a memory
  • a useful truth
  • a follow up question

Reject questions that only collect data.

Premium Test

Reject the question if it does not feel worth paying for.

A premium question should feel:

  • specific
  • warm
  • memorable
  • useful
  • fun
  • relationship focused

Option Quality

For multi_choice and single_choice:

  • use 4 to 6 options
  • avoid obvious correct answers
  • avoid overlapping options
  • avoid vague options like "other"
  • avoid all negative options
  • avoid options that shame either partner
  • keep options short
  • make options feel realistic
  • make every option answer the exact prompt

Bad options:

  • Be better
  • Communicate more
  • Fix us
  • Other

Good options:

  • Make dinner easy
  • Take a short walk
  • Send a sweet text
  • Save me the good blanket

Written Question Gate

A written question is allowed only when typing creates more value than choices.

Reject written questions that ask for:

  • basic preference
  • yes or no
  • generic reflection
  • long emotional labor
  • forced vulnerability

Written questions should create stories, memories, or meaningful appreciation.

Scale Question Gate

Use scale only when intensity matters.

Good scale uses:

  • How close does this feel?
  • How comfortable are you?
  • How often does this happen?
  • How important is this?
  • How ready are we?

Bad scale uses:

  • random preferences
  • jokes
  • simple choices
  • topics better served by options

Emotional Safety

Reject questions that pressure users to:

  • confess
  • forgive
  • reveal trauma
  • admit guilt
  • justify boundaries
  • defend needs
  • disclose sexual details before consent context
  • compare partners cruelly

The app should open doors, not corner people.

Fun Requirement

Every category must include playful moments.

Even serious categories need:

  • small humor
  • soft wording
  • hopeful prompts
  • low pressure options
  • practical repair ideas

No category should feel like 250 tiny court summons.

Category Fit

Every question must clearly belong to its category.

Reject questions that could fit anywhere.

Bad for Gratitude:

What do you like most about us?

Better:

What's one small thing I do that deserves more credit?

Relationship Anchor

Every question should connect to the couple.

Prefer:

  • us
  • we
  • you and me
  • our life
  • our future
  • our memories

Avoid generic survey voice.

Final Human Read

Before committing a pack, read 30 random questions aloud.

Reject the pack if it sounds:

  • robotic
  • repetitive
  • clinical
  • boring
  • too similar
  • too heavy
  • too generic
  • like a wellness worksheet
  • like a couples counselor wrote it

Final Approval

A pack is ready only when:

  • schema passes
  • counts pass
  • duplicate checks pass
  • tone review passes
  • category fit passes
  • consumer test passes
  • no obvious AI patterns remain